A user-wearable electronic device includes a housing configured to be worn on a user's torso, a plurality of sensors disposed in the housing, including a first sensor to sense motion of the user and produce raw activities of daily living (ADL) data, and a biometric sensor to sense one or more biometric characteristics of the user. One or more processors in the electronic device or an intermediary device generate, for a sequence of time periods, ADL identification information by processing the raw ADL data using one or more neural networks pre-trained to recognize a predefined set of ADLs. Each pre-trained neural network includes a plurality of neural network layers, including at least one layer that includes a recurrent neural network. Reports that include ADL information corresponding to the generated ADL identification information for time periods in the sequence of time periods are transmitted to a monitoring system.
